Claims

Factual · Science and research methods

“The idea is rooted in Hubel and Wiesel’s classic 1962 work on the cat’s primary visual cortex.”

Yann LeCun, co-author of paper · 2010 · Convolutional Networks and Applications in Vision (yann.lecun.com)

The idea behind multi-stage Hubel-Wiesel architectures is rooted in Hubel and Wiesel's 1962 work on the cat's primary visual cortex.

Resolved true The record shows it happened or was so.

The second judge would not resolve this claim; the verdict shown is the first judge's.

What would count as true or false

True if Hubel and Wiesel's classic paper on the cat's primary visual cortex (receptive fields, simple and complex cells) was published in 1962 and is recognised as the inspiration for multi-stage architectures that alternate filtering and pooling (Neocognitron, ConvNets). Mostly true if the lineage is correct but the cited year or detail is slightly off (for example their key work being dated 1959 or 1968). Mixed if the 1962 paper exists but its role as the root of these architectures is disputed. Mostly false if the architectures' origin is credited mainly to other work and the Hubel-Wiesel link is minor. False if no such 1962 work on the cat visual cortex exists, or there is no link.

Evidence

1. Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position Biological Cybernetics (K. Fukushima, NHK Broadcasting Science Research Laboratories) · 1980

   “After completion of self-organization, the network acquires a structure similar to the hierarchy model of the visual nervous system proposed by Hubel and Wiesel (1962, 1965).”

2. Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position Biological Cybernetics (K. Fukushima, NHK Broadcasting Science Research Laboratories) · 1980

   “It is true that the hierarchy model by Hubel and Wiesel does not hold in its original form.”

3. Who invented convolutional neural networks? IDSIA (Jürgen Schmidhuber) · 7 December 2024

   “★ 1979: Kunihiko Fukushima published the basic CNN architecture with convolution layers and downsampling layers,[CN79][CN80][CN21] inspired by neurophysiological findings of Hubel and Wiesel.[HUW59-68]”

4. Neocognitron - Wikipedia Wikipedia · 24 July 2006

   “The neocognitron was inspired by the model proposed by Hubel & Wiesel in 1959.”

Reasoning

Fukushima's 1980 Neocognitron paper (E1) says the self-organised network takes on a structure similar to "the hierarchy model of the visual nervous system proposed by Hubel and Wiesel (1962, 1965)". That supports both the 1962 date and the link to multi-stage architectures. Schmidhuber's history of CNNs (E3) also says Fukushima's architecture of convolution and downsampling layers was inspired by Hubel and Wiesel's findings, though it cites them only as "HUW59-68" and gives no single year. E4 dates the inspiring model to 1959, a variation in year that the criteria list as an example for mostly true, while E1 cites the Hubel and Wiesel model as 1962 and 1965. The cat detail is not stated in these quotes, and E2 says the biological model does not hold in its original form. Neither point disputes that the architectures descend from Hubel and Wiesel's work.

Cite this claim

Yann LeCun, “The idea behind multi-stage Hubel-Wiesel architectures is rooted in Hubel and Wiesel's 1962 work on the cat's primary visual cortex.” (said 2010). Verdict: resolved true. Reliability API, claim c_47bceb0087f6, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_47bceb0087f6